Robots are becoming increasingly integrated into everyday environments — from homes and hospitals to factories and farms. However, safely operating around humans requires more than strength or speed. Robots must also sense their surroundings, detect physical contact, and respond quickly. Conventional sensors, especially those embedded in soft materials, often fall short when it comes to real-time, large-area tactile and proximity sensing. To address this challenge, a research team led by professor Van Anh Ho and Quan Khanh Luu from Japan Advanced Institute of Science and Technology (JAIST) developed ProTac, a novel vision-based soft sensing skin for robots. According to the researchers, the work represents a scalable approach to multimodal robotic perception. Conceptual overview of the vision-based ProTac sensing technology. ProTac can actively switch between proximity and tactile sensing modes, relying on input images captured by inner cameras and a soft functional skin with controllable transparency. Courtesy of JAIST/Van Anh Ho. At the core of ProTac is a polymer-dispersed liquid crystal (PDLC) layer that can switch between transparent and opaque states when voltage is applied. When transparent, embedded cameras can “see” through the skin to detect nearby objects. When opaque, those same cameras track how the skin deforms, allowing the robot to sense touch, pressure, and the location of contact. This dual-mode sensing enables real-time environmental perception using a single soft layer, eliminating the need for complex embedded electronics. “This innovation with sensor design simplicity allows robots to perceive both contact and nearby obstacles across a large-area skin in real time, which is difficult to achieve with conventional electronic skins,” Ho said. To validate the design, the team developed a prototype called the ProTac link, a cylindrical robot arm segment wrapped in a soft sensing skin and equipped with stereo cameras at both ends. This prototype can detect approaching objects from multiple angles, estimate distance, and recognize multiple touch points with high accuracy. It also supports adaptive behaviors such as proximity-based speed adjustment and reflexive contact avoidance. “ProTac can be applied to dexterous robotic manipulation in various domains where safety and delicate physical interaction are critical,” said Luu. The researchers envision broad applications across collaborative robotics, assistive devices, and soft robotic systems in agriculture or domestic care. In factories, ProTac-equipped robots could slow or stop when a human worker approaches. In elder care, service robots could provide physical assistance while maintaining safe, responsive contact. “In the long term, this research lays the groundwork for intelligent general-purpose humanoid robots equipped with full-body multimodal soft sensing skins, especially those with a sense of touch,” said Ho. Alongside the skin, the team developed learning algorithms and control strategies that convert sensor data into robotic actions, enabling responsive and autonomous behavior in dynamic environments. Designed with simplicity, modularity, and scalability in mind, the ProTac system can be integrated into both new and existing robotic platforms. To accelerate progress in the field, the researchers have made their design files, models, and software open source. By combining mechanical simplicity with multimodal perception, ProTac brings robots a step closer to understanding — and safely interacting with — the complex, tactile world around them. The research was published in IEEE Transactions on Robotics (www.doi.org/10.1109/TRO.2025.3593087).